Lightning-AI/pytorch-lightning

Resuming should allow to differentiate what to resume (steps/opti/weights)

開放

#5,339 建立於 2021年1月3日

 (26 則留言) (18 個反應) (0 位負責人)Python (3,233 個分叉)batch import
featurehelp wantedpriority: 1

倉庫指標

星標
 (26,687 顆星)
PR 合併指標
 (平均合併 9天 15小時) (30 天內合併 3 個 PR)

描述

Currently it is possible to either resume only the full training state (epoch/global steps / optimizer / scheduler options / and weights), or only the weights.

I would like to be able to switch the optimizer at some point, i.e. skip restoring optimizer/scheduler, but still load the epoch/global steps. I only see a way to do this with hacks at the moment. Any other way? Could this be a feature, to specify in the trainer.init function specifically what to restore/what not to restore?

貢獻者指南